MIMO systems and Spatial Diversity. - Graz University of ... · MIMO systems and Spatial Diversity. ... Wireless communication and MIMO systems. ... Extended channels. MIMO systems

MIMO systems andSpatial Diversity.

Dmitriy Shutin

[email protected].

Signal Processing and Speech Communication Laboratory

www.spsc.tugraz.at

Graz University of Technology

MIMO systems andSpatial Diversity. – p.1/36

Content of the talk

PART I.Wireless communication and MIMO systems.Wireless channels.

Part II.Spatial Diversity


Outline – PART I.

PART I.What is MIMO

Multiple antennas in communication

Channel and channel models.Channel and signal models.


PART II.


Outline – PART II.

What is diversity?Diversity schemes.What do we actually win.

Implementing diversity – IID channelsRX diversity(SIMO).Alamouti scheme and TX diversity (MISO).RX-TX diversity (MIMO).

Extended channels.


What is diversity?Diversity – Multiple independent look at the same transmittedsignal.


What is diversity?Fading impairs wireless link.

Use multiple antennas to create alternative transmissionbranches.

216 218 220 222 224 226

−160

−140

−120

−100

−80

−60

distance, λ

Pow

er v

aria

tions

, dB

1st receiver antenna2nd receiver antenna


What is diversity?If p = P (signal is in fade)

Having M independent branches results

P (all M branches are in fade) = pM � p

There areFrequency diversityTime diversitySpace diversity


Diversity schemes.Space diversity: RX/TX has multiple antennas, spaced dmeters apart such that

d > Dc

Frequency diversity: signal is transmitted over two carrierfrequencies, spaced ∆f , such that

∆f > Bc ∼1

RMS delay spread

Time diversity: the same signal is re-transmitted with adelay ∆t such that

∆t > Tc =1

BD


Diversity GainTo exploit diversity we need:

Combine multiple branches in some way.

Make sure branches are independent. Separation> Bc, Tc, Dc

Signal model for each branch:

yi =

√

Es

Mhis + ni, n = 1, . . . ,M

hi – flat-fading channel coefficient;Es – symbol energy;

ni – additive noise.


Diversity Gain, cont’dTo combine branches – Maximum Ratio Combining:

z =M∑

i=1

h∗i yi

Each branch increases SNR

η =1

M

M∑

i=1

(

|hi|2 ·Es

N0

)

= α · ρ

α =1

M

M∑

i=1

(

|hi|2)

– branches gain, ρ =Es

N0

– Single-branch SNR


Diversity Gain, cont’dAssuming ML detection, averaged probability of the symbolerror can be upper-bounded as

P e ≤ Ne

(

ρd2min

4M

)−M

Ne - number of nearest neighborsd2

min – minimum separation of the symbol constellation.

ρ = Es

N0– averaged SNR at the receiver antenna.


Diversity Gain, cont’dEffect of diversity on the SER performance in fading channels.

Diversity affects the slope of the SER-SNR curve.MIMO systems andSpatial Diversity. – p.13/36

Coding gain vs diversity gain

Approximate P e can beexpressed

P e ≈c

(γcρ)M

c – modulation/channel constantγc ≥ 1 – coding gainM – diversity order


Spatial vs Time/Frequency diversity

Spatial diversityNo additional bandwidth requiredIncrease of average SNR is possibleAdditional array gain is possible

Time/Frequency diversityTime/frequency is sacrificedNo array gain.Averaged receive SNR remains as that for AWGNchannel.


Implementing Spacial diversity

Receiver diversity – SIMO

Transmit diversity– MISO

RX/TX diversity – MIMO


Receive antenna diversity

Let us assume flat fading

y =√

Eshs + n, h = [h1, . . . , hMR]T

Maximum Ratio Combining at the receiver

z = hHy =√

EshHhs + hHn

MRC assumes perfect channel knowledge at RX.


Receive antenna diversity, cont’d

Averaged probability of symbol error is given as (at highSNR regime and rich scattering)

P e ≥ Ne

(

ρd2min

4

)−MR

Averaged SNR at the receiver

η = MR · ρ

Array gain is MR (10 log10 MR [dB])


Receive antenna diversity - performance

Can be better than AWGNdue to array gain

For BER> 10−5

outperforms AWGN dueto array gain.

RX diversity extracts full diversity and array gain!


Transmit antenna diversityStraightforward approach is useless:

Signal model (for MT = 2); h1, h2 ∼ N (0, 1)

y =

√

Es

2(h1 + h2)s + n

Equivalently, h ≡ (h1 + h2)/√

2, and h ∼ N (0, 1)

y =√

Eshs + n

Unlike RX diversity, pre-processing is needed.


TX diversity – Alamouti(MISO)Preprocessing without channel knowledge.


Alamouti scheme, cont’d

Channel is constant over two symbols intervals andflat-fading

h = [h1, h2]

Two received symbols are:

y1 =

√

Es

2h1s1 +

√

Es

2h2s2 + n1

y2 = −√

Es

2h1s

∗2 +

√

Es

2h2s

∗1 + n2


Alamouti scheme, cont’d

Now, received vector is formed as y = [y1, y∗2]T

y =

√

Es

2

[

h1 h2

h∗2

−h∗1

][

s1

s2

]

+

[

n1

n∗2

]

=

√

Es

2Heffs + n

MRC gives ( using HHeff

Heff = ‖h‖2I)

z = HHeffy =

√

Es

2‖h‖2

Is + n

which simplifies to zi =√

Es

2‖h‖2si + ni.


Alamouti scheme – Performance

High SNR regime gives

P e ≤ Ne

(

ρd2min

4 · 2

)−2

No array gain: η = ρ

Full transmit diversity forMR = 2

How to implement array gain?


MISO with known channel

Flat-fading channel is given as h = [h1, . . . , hMT]T

Signal at the receiver

y =

√

ES

MT

hHws + n

where w =√

MTh

‖h‖

This scheme is known as transmit MRC.


Transmit MRC – Performance

High SNR regime gives

P e ≤ Ne

(

ρd2min

4

)−MT

Array gain in richscattering η = MT · ρEquivalent to RX diversity.

Transmitter should be aware of the channel.


Alamouti scheme & MIMO

Transmit symbols like MISO Alamouti

Flat-fading channel matrix

H =

[

h1,1 h1,2

h2,1 h2,2

]

Receive symbols (MR = 2)

y1 =

√

Es

2H

[

s1

s2

]

+ n1, y2 =

√

Es

2H

[

−s∗2

s∗1

]

+ n2,

equivalently...


Alamouti scheme & MIMO, cont’d

By defining y = [y1,y∗2]T

y =

√

Es

2

h1,1 h1,2

h2,1 h2,2

h∗1,2 −h∗

1,1

h∗2,2 −h∗

2,1

[

s1

s2

]

+

[

n1

n2

]

MRC is formed (using HHeff

Heff = ‖H‖2F I)

z =

√

Es

2‖H‖2

F Is + n, or,

zi =

√

Es

2‖H‖2

F si + ni


Alamouti scheme & MIMO, cont’d

High SNR regime gives and rich scattering

P e ≤ Ne

(

ρd2min

4MT

)−MRMT

Diversity order MRMT

Array gain η = MRρ

Thus, without channel knowledge only array gain is obtained.


MIMO with channel knowledge

If channel known – transmit MRC is used

y =

√

Es

MT

HHws + n

Receiver forms weighted sum z = gHy

Vectors w and g are singular vectors from SVDdecomposition of H

H = UΣVH ; w ∈ U, g ∈ V

w and g correspond to the max singular value σ2max of H

This is known as dominant eigenmode transmission – DET


MIMO with DET – Performance

Diversity order MRMT

SNR η = E{σ2max}ρ

max(MT ,MR) ≤ E{σ2max}

E{σ2max} ≤ MT MR

Channel knowledge increases array gain.


Summary

Configuration Array gain Diversity gainSIMO (Ch.Un.) MR MR

SIMO (Ch.Kn.) MR MR

MISO (Ch.Un.) 1 MT

MISO (Ch.Kn.) MT MT

MIMO (Ch.Un.) MR MRMT

MIMO (Ch.Kn.) max(MT ,MR) ≤ E{σ2max},

E{σ2max} ≤ MTMR

MRMT


Diversity order and channel variability

Can be quantified as

µvar =1√

MRMT

Link stabilizes asµvar → ∞


Diversity order for extended channels

Elements in channel matrix H areCorrelatedHave gain imbalancesFade with Ricean statistics

Highest possible diversity −→ rich scattering (IID elementsin H)

Let as consider Alamouti scheme with MT = MR = 2


Correlations in H

Diversity order decreasesto r(R), where R is (4 × 4)covariance Matrix:

R = E{vec(H)vec(H)H}

Rank r(R) = 1 −→ fullcorrelation.

No spatial diversity.

Array gain is present.


Ricean fading of H

Ricean fading occurswhen LOS is present.

As K → ∞ link stabilizes.K-factor is given as

K =A2

LOS

σ2

diffuse


MIMO systems and Spatial Diversity. - Graz University of ... · MIMO systems and Spatial Diversity. ... Wireless communication and MIMO systems. ... Extended channels. MIMO systems

Documents